Scattered
A request arrives in a channel. The ticket lives in a tracker. The follow-up sits in one person's inbox. The review is a calendar invite. Each tool is fine. Nobody holds the whole picture, so the thing that matters is the thing nobody saw.
Private early access
Your work is scattered: the request is in a Slack thread, the ticket is in Jira, the reply you owe is in your inbox, the review is on tomorrow's calendar. Alfred reads all of it and tells you the few things it can finish for you — written out in full, waiting for you to press yes.
No spam, no newsletter. One email when it's your turn, and one question we'd genuinely like answered.
The demo
Being recorded now.
Two minutes, start to finish: a message arrives, the ticket and the reply get drafted, one click does both. Join the waitlist and it will be the first thing we send you.
The problem
Not a software problem, exactly. Three habits that come free with every tool you buy.
A request arrives in a channel. The ticket lives in a tracker. The follow-up sits in one person's inbox. The review is a calendar invite. Each tool is fine. Nobody holds the whole picture, so the thing that matters is the thing nobody saw.
Someone reads the thread, opens the tracker, writes the issue, goes back and writes the reply, then builds the deck the morning of the review. Every hand-off costs detail and most of an hour, and it is the same hand-off every week.
The AI we have now is good at answering and poor at noticing. It acts when told, and it has no idea who is allowed to see what — so the work is either missed or it leaks. Neither is a thing you can build a week around.
How it works
Four steps, and you are only in the last one.
Alfred reads your channels, your tracker, your mail and your calendar. What arrives files itself against the customer, account or project it names, so the context is already in place before anyone asks for it.
It works out what is actually waiting, whose it is, how urgent it is and what it is worth — and it looks things up before it answers rather than guessing from the last thing it read.
It writes the work out in full on the right person's list: the ticket, the reply, the task, the brief for tomorrow's meeting. Drafted and editable. Nothing has left the building yet.
You read it, change a word if you like, and press yes. Alfred runs the steps as you — your name on the ticket, your voice in the reply — and writes every one of them to an audit log.
One ordinary morning
The ordinary case, start to finish. Nothing here is a mock-up; it is the path the product takes today.
In a shared channel, a customer's engineer writes: “SSO fails after 2.4.” No form, no ticket, no handover note. Just a sentence in a thread.
It works out which account this is, drafts the ticket, and drafts the reply twice — once with the ticket link, once without, because sometimes you answer before you file.
The person who owns that account, with what they need beside it: the open tickets, the renewal date, what the account is worth. Private to them.
The ticket appears in the tracker under their name. The reply posts in the thread as them, with the link. Both steps are in the audit log, and both can be undone.
Measured in production: from the message arriving to the drafted ticket and reply, 1.2 to 1.4 seconds.
Built and running
Built and running, not a roadmap. We would rather show you a short list that is true.
Your to-dos, the things waiting on your approval, and the things Alfred is offering to do — in the order they actually matter.
Ask about an account, a project or a thread. It searches and looks things up before answering, and tells you what it read.
The ticket, the reply, the email you owe: written out, editable, approved one by one or all at once, and undoable after.
The replies, tickets, calls and threads nobody has answered yet, in one place, with a draft already attached to most of them.
Save a meeting and Alfred offers the brief or the deck for it there and then — the account's health, renewals and risks, ready before the call.
Your own visibility is applied to every search, answer and action. Alfred cannot ground an answer in something you are not allowed to open.
Connections
Alfred is built so that a connection is just a connection. The order we build the next ones in is the order you ask for them — which is one of the questions we ask after you join.
Every channel Alfred is invited to
Issues created and commented as you
Gmail and Calendar
Asked for most, not built yet
Proof
Claims are cheap, so here are the numbers we actually measure, and the one thing we haven't earned yet.
Message to drafted ticket and reply
Measured on production traffic, not on a laptop.
Answer quality, graded independently
A fixed set of hard questions — facts, privacy, refusals, made-up answers — marked after every release. It was 5 of 16 on the first run.
Automated tests on every change
About 7,900 on the server and 7,100 in the interface, including one that plants a secret in someone else's accounts and proves it never reaches your answers.
Written to an audit log
Who, what, when and as whom — for approvals and refusals alike.
SOC 2 controls are written into the code and checked by the build, which fails if one slips. The audit itself is ahead of us, not behind us — so we are not going to put a badge on this page. Ask us where we are and we will tell you plainly.
Who it's for
If your work arrives as messages and leaves as tickets, replies and documents, it is for you.
Escalations, renewals, the health of a book, the brief you needed ten minutes before the call.
Sales, support, operations, engineering. The tools differ; the copying by hand does not.
No customer-success function, no time to switch tabs, and everything on one person's shoulders.
What is waiting across the team, what is overdue, what moved this week and who to ask about it.
FAQ
Alfred is an agentic to-do list for business teams. It connects to the tools your work already lives in — chat, your issue tracker, mail and calendar — reads what is waiting, and proposes the next action as a finished draft: a ticket, a reply, a task, a document. A person approves it in one click, and Alfred carries it out as them.
A to-do app holds what you remembered to type into it. Alfred finds the work itself, in the places it actually arrived, and writes the entry for you — with the ticket and the reply already drafted underneath it.
An assistant waits to be asked, and answers. Alfred notices, and offers to act. It goes first, with the work already written out, and it knows who is allowed to see what — which an assistant you paste context into cannot.
No. Nothing leaves your company without a person pressing yes. Alfred drafts, you read it, you can edit it, you approve step by step or all at once — and you can undo it afterwards. Every action, and every refusal, is in the audit log.
No, and this is the part we have tested hardest. Your own visibility — your role, your department, where you sit in the org chart — is applied to every search, every answer and every action. There is a test that plants a secret in a colleague's accounts and proves it never reaches your answers, your search results or the API.
Slack, Jira, and Google Workspace — Gmail and Calendar. More are coming, and the order is decided by what people on the waitlist tell us they use. Teams, Outlook, Linear, Zendesk, Salesforce, HubSpot, Notion and Asana are the ones asked for most so far.
Alfred is in private early access and we are letting people in a few at a time, so that we can sit with each team rather than watch a dashboard. Joining the waitlist puts you in that queue; answering the questions afterwards moves you up it, honestly.
We have not set pricing, and we would rather not invent a number to fill this space. It is one of the things we are asking early teams about.
Join the waitlist, watch the two-minute demo, and tell us what you would want it to do first. We read every answer.